Monday, November 7, 2016

99.9% vs 66.9%

Today — the day before the election — the polling oracles of Sam Wang at the Princeton Election Consortium and Nate Silver at FiveThirtyEight are showing two very different predictions.  PEC says Hillary Clinton has a 99.9% chance of winning while 538 is saying it’s 66.9%.  Why?


Princeton Election Consortium 11-07-16

Five Thirty Eight 11-07-16

Five Thirty Eight 11-07-16

At the moment 538 isn’t talking, but Sam offers an explanation.  Fair warning: it’s long, detailed, and gets into statistical interpretation that will test your wonk tolerance, but it’s worth the read to see why two different organizations without — at least on the surface — an agenda for one candidate over another can come to two very different results when they’re looking at the same data.

And for a happy medium, there’s the Upshot at the New York Times, which, with Hillary Clinton at 84%, lands exactly in the middle between PEC and 538.

Take your pick.

2 barks and woofs on “99.9% vs 66.9%

  1. I direct your attention to a post on Huffington written by Evan Cohan of the Brattle Group, whatever that is, titled “What’s The Matter With 538?” in which he tears apart Nate Cohen’s methodology and assumptions. The penetrating analysis is way over my head, but I’m impressed by the conclusions. I’ll never obsess over 538 again. I now believe you simply can’t trust it.

Comments are closed.